February 2021


Salman Ullah

Why quality of deal flow is so critical in venture capital – a scientific approach

No doubt many of you have been following the vicissitudes of Covid testing in the US. Apart from the bottlenecks in actually accessing the tests, there’s also the ever-present challenge of test accuracy. In other words, if you test positive, are you really positive? And, if you test negative are you really negative? Most Covid tests claim an accuracy of 95% (strictly speaking there are two measures of accuracy: selectivity and sensitivity; more on this later) which would suggest that they are, in fact, very useful tests. At this point, the Reverend Thomas Bayes may raise his hand to point out that testing positive means that your likelihood of being actually positive is only about 50%. Why is that? And what does any of this have to do with venture investing?

Let’s tackle the first question. An important factor in understanding why Covid testing is only about 50% accurate is understanding the underlying actual infection rate. At present, the infection rate is only about 5%. So, out of a population of 1,000, only 50 will be infected and 950 will be virus free. However, the test will give a false positive in 5% of 950 cases. In other words, the test will incorrectly identify 48 people as being positive -- the ratio of actual positives to tested positives is 50/(50+48) or about 50%.

If you run through the math again with a much higher underlying infection rate of 25%, you’ll find the test reveals that a person who tests positive is about 86% likely to actually be positive. The takeaway is that the higher the underlying infection rate, the better the test is at finding people who are actually infected.

The math is captured concisely by Bayes’ Theorem - see the box for details.

Bayes’ Theorem

P(A|B) = [P(B|A) * P(A)/P(B)

Here, P(A|B)= probability of A given B and P(A) is the probability of A. (Similarly for the other terms.)

For doing the calculations, one more relation is helpful:

P(B) = P(B}A) * P(A) + P(B|Not A) * (1-P(A))

The analysis we just went through can be replicated, essentially, word-for-word to the case of venture investing, where we are trying to pick the winners from the losers based on our judgement. The following table captures the key inputs:

Next, we’ll make a simplifying assumption around our judgement: we are just as good at rejecting losers as we are at selecting winners, i.e. sensitivity equals selectivity.

We’ll run the calculation for the probability of successful selection for different values of our judgement expressed as a probability (where, for example, 95% means we make a mistake only 5% of the time) and quality of pipeline (where, for example, 2% means that 2% of the companies we see are objectively good companies).

The body of the table displays the ultimate likelihood of picking a winner for a given quality of leads and a given quality of judgement.

For example, looking across the first row with 2% lead quality, we see that if our innate ability is to be correct 90% of the time, then we ultimately end up picking a winner only 16% of the time. Our judgement has to increase to about 98% before we can expect to be right 50% of the time. If we want to be right at least 50% of the time, our required judgement needs to be only 95% for the case where 5% of the leads are good, and a mere 90% if 10% of leads are of high quality.

Finally, let’s take the last column with a judgement score of 98% (possibly an aspirational target at best). Notice how as the lead quality increases there’s a massive jump in overall success rate from a low of 50% at 2% lead quality, to all the way to 84% for 10% lead quality.

So, while the precise values of the inputs are themselves subjective, the conclusion is clear: both judgement and quality of deal flow are important, however, it’s the latter that is most important.

If the underlying quality is low then we’ll end up investing in too many false positives. Venture is not primarily a volume game when it comes to deal flow. If anything, too much unfiltered volume will drive too many false positives (again, no one is infallible) and hurt returns. Scouring the unfiltered universe of startups will not drive great returns.

Instead, it really is all about the ‘Glengarry leads’, and the best source of these high-quality leads are portfolio company referrals and introductions from executives in our network. We strive to cultivate these two network nodes because their referrals drive our greatest returns. It’s just math.